Post 9uhO3CU2ElXdMoCQca by chloekek@floppy.tokyo
 (DIR) More posts by chloekek@floppy.tokyo
 (DIR) Post #9ughFvxEht5zq3gQtc by sir@cmpwn.com
       2020-05-04T01:31:59Z
       
       2 likes, 3 repeats
       
       The ONLY way to make reliable, working software is to make it simple.
       
 (DIR) Post #9ugiV3thYbeIlso8IK by middlepath@mastodon.social
       2020-05-04T01:44:12Z
       
       0 likes, 0 repeats
       
       @sir I think people confuse the terms complicated and complex. The former is unmaintainable spaghetti, the latter is just such a hard problem to solve that it *requires* a simple solution.X11 is very complicated, wayland is very complex ;)
       
 (DIR) Post #9ugiorcotXnkV8LzTE by sir@cmpwn.com
       2020-05-04T01:44:59Z
       
       0 likes, 0 repeats
       
       @middlepath I think wayland on the whole is pretty simple, at least for what it is
       
 (DIR) Post #9ugjZ5XbUKYWytxIEy by middlepath@mastodon.social
       2020-05-04T01:57:48Z
       
       0 likes, 0 repeats
       
       @sir totally my point.It’s a simple design to a very complex problem. But it’s not “simple” in the sense that you can keep all the code on your screen at once.
       
 (DIR) Post #9ugjg7e43ymtXCKbqK by sir@cmpwn.com
       2020-05-04T01:58:25Z
       
       0 likes, 0 repeats
       
       @middlepath I think the amount of code you can keep on your screen at once has little to do with being complex or complicated
       
 (DIR) Post #9ugmiRH4Ly4Ir3N72G by riking@social.wxcafe.net
       2020-05-04T02:32:49Z
       
       0 likes, 0 repeats
       
       @sir _Building Secure Reliable Systems (2020)_ agrees with you.
       
 (DIR) Post #9uh6qAorPlSjDj8vKK by wolf480pl@mstdn.io
       2020-05-04T06:19:22Z
       
       0 likes, 0 repeats
       
       @sirImplementation simplicity is slightly more important than interface simplicity, right?
       
 (DIR) Post #9uh7QBd1spoA9fmVP6 by lanodan@queer.hacktivis.me
       2020-05-04T06:26:09.612837Z
       
       0 likes, 0 repeats
       
       @wolf480pl @sir Both are linked, be it an interface for humans or for programs, if the interface is complicated then so is the rendering. Better simplify the interface actually, that can leave room for more implementation simplicity.
       
 (DIR) Post #9uh8FqF0e8lEsiB39U by wolf480pl@mstdn.io
       2020-05-04T06:35:12Z
       
       0 likes, 0 repeats
       
       @lanodanLet me rephrase the question: should an interface be a full abstraction layer, trying to hide underlaying quirks from whoever uses it, or should it expose those quirks?IOW, EINTR or not EINTR?@sir
       
 (DIR) Post #9uh8seBfOdgqn57Lyi by lanodan@queer.hacktivis.me
       2020-05-04T06:42:30.481816Z
       
       0 likes, 0 repeats
       
       @wolf480pl @sir It should expose them, if you hide quirks or imperfections under the carpet they'll eventually surface in even worse ways.EINTR → Not sure but reading about it reminds me of how some shitty code doesn't handle malloc returning NULL. I'd rather have shitty code crashing than trying to avoid it (overcommitting is a huge mess for reliability).
       
 (DIR) Post #9uhBGmYw6nH9cmdcJ6 by rick_777@cybre.space
       2020-05-04T07:08:58Z
       
       0 likes, 0 repeats
       
       @wolf480plIf the implementation can't fix those quirks, they should be exposed so that the code using that interface can work around them. But only as a last resort.@lanodan @sir
       
 (DIR) Post #9uhBLDz06OWbpFwoc4 by wolf480pl@mstdn.io
       2020-05-04T07:09:48Z
       
       0 likes, 0 repeats
       
       @rick_777 @lanodan @sir what if those quirks can be fixed by increasing the implementation complexity three times?
       
 (DIR) Post #9uhDdJWESSGE7GFEuG by lanodan@queer.hacktivis.me
       2020-05-04T07:35:45.741011Z
       
       0 likes, 0 repeats
       
       @wolf480pl @rick_777 @sir Depends on the size of the implementation, something barebones times three is still nicely maintainable, but if you pick something already quite bloated to begin with you're going to have issues, specially as adding code adds bugs.Also workarounds are still the same "stashing under the carpet" thing, like if malloc or a syscall can't fail without crashing your program you need a reality check, hardware has limits and will fail and you need to acknowledge it. Having a warning or an error message doesn't always means a bug.
       
 (DIR) Post #9uhGeooFuB37JZ5HSC by wolf480pl@mstdn.io
       2020-05-04T08:09:22Z
       
       0 likes, 0 repeats
       
       @lanodan @sir @rick_777 Ok, but AFAIK nobody is speaking about workarounds, it's more about pushing the edge cases onto the user of the API.
       
 (DIR) Post #9uhO3CU2ElXdMoCQca by chloekek@floppy.tokyo
       2020-05-04T09:30:58Z
       
       1 likes, 0 repeats
       
       @sir This is one of those popular things to say that don’t actually say anything.
       
 (DIR) Post #9uhPwkCIgtl1oHxAo4 by stman@mastodon.social
       2020-05-04T09:52:31Z
       
       0 likes, 0 repeats
       
       @sir Fully agree, but adding this :* for the moment. * until better strategies or software design technics are invented.But simplicity will alway be a key ingredient.